With　the　continuous　improvement　of　China＇s　broadband　optical　network　capabilities,　the　focus　of　users＇　attention　has　shifted　to　application　perception.　The　problem　of　frequent　broadband　disconnection　has　long　plagued　first-line　units　and　greatly　affected　users＇　trust　in　broadband　quality.　Aiming　at　the　problem　of　long　processing　time　and　difficult　location　of　boardband　frequent　offline　fault,　this　paper　proposes　a　Hadoop-based　fault　port　detection　scheme.　Hadoop　distributed　file　system　is　used　to　store　the　user＇s　online　and　offline　data,　and　MapReduce　is　used　for　parallel　calculation.　Then,　the　fault　port　is　located　through　the　correlation　of　time　series.　Finally,　the　data　is　transmitted　to　the　database　of　the　display　system　through　Sqoop　In　the　display　system,　the　fault　situation　of　the　whole　network　users　is　analyzed.　Frequent　offline　failures　change　from　the　original　user＇s　application　to　community　maintenance　to　the　detection　platform＇s　self　discovery　of　fault　ports,　which　is　a　successful　practice　of　cloud　network　integration　big　data　operation,　and　also　reduces　the　burden　of　operation　and　maintenance　personnel.　©　2021　IEEE.